Estimation of the channel conditions between the transmitter and receiver is a necessary step for many communications systems to enable detection and optimal processing of a data stream received from a signal source. So as to enable the necessary channel estimations, most of these systems embed reference symbols in the data stream that are known a priori to the receiver.
In many cases, joint channel estimation techniques are necessitated by the presence of multiple, simultaneous data streams received from multiple signal sources, for example due to multiple access interference (i.e. a plurality of users 100 transmitting a plurality of signals 102 to a base station 104 over the same communication channels, as illustrated in FIG. 1) and/or Multiple Input Multiple Output (“MIMO”) multi-path propagation. The need for joint channel estimation can be even greater in a Centralized Radio Access Network (“C-RAN”). These joint channel estimation techniques attempt to obtain multiple channel estimates simultaneously from multiple sources, thereby improving the channel estimation accuracy when the reference symbols from multiple signal sources are not orthogonal.
Computationally efficient linear methods for joint channel estimation are well known, such as the least squares filter or the minimum mean square error filter. These techniques attempt to obtain good channel estimates through modeling the cross correlation between the reference symbols, and by creating a filter to separate them so as to improve the accuracy of the estimates. These filters are computationally simple, and perform adequately in many circumstances.
Typically, the number of reference symbol samples needed for joint channel estimation must be greater than the number of channel estimates to be computed. One approach is to include a plurality of time-separated reference symbol samples in the calculation, under the assumption that the channels are static over relatively short periods of time. In addition, with reference to FIG. 2, in the case of frequency-multiplexed data transmissions, such as Orthagonal Frequency Division Multiplexing (“OFDM”) or Single Channel—Frequency Domain Multiple Access (“SC-FDMA”) 202 encoded transmissions, in which each data stream of symbols 200 is split into a plurality of sub-streams that are converted to a time domain signal by an Inverse Fast Fourier Transform (“IFFT”) 206 and transmitted over frequency-separated “subcarriers,” a plurality of reference symbol samples that are frequency-separated across groups of subcarriers can be included in the joint channel estimation. This approach assumes that the channels are “static” over a frequency bandwidth that encompasses a plurality of subcarriers. Note that some or all of the signal sources may not use all of the a subcarriers that are included in the data transmission.
Accordingly, in order to obtain the dimensionality necessary to distinguish a plurality of signals and obtain high quality channel estimates for each of the signal sources, joint channel estimation is often performed over a time/frequency sampling “window” containing reference symbols that are separated from each other in time and/or in frequency. Implicit in this approach is an assumption that the channels do not change significantly over the time dimension and/or the frequency dimension of the “window.” This is referred to as the “static channel assumption.” Even when the channels are not fully static in time or frequency, this assumption is often sufficiently valid to enable joint channel estimation over a relatively small time/frequency sampling window. However, various issues can arise due to invalidity of these static assumptions, and/or because it is not guaranteed that sufficient dimensionality will be obtained even if the number of reference symbol samples used in the calculation exceeds the number of channel estimates, due to excess correlation of the reference symbol samples and/or of the channels.
With reference to FIG. 3, after receiving signals containing raw time domain data 300 and performing a Fast Fourier transform (“FFT”) 302 to separate the subcarriers 304, joint channel estimation 306 of frequency-multiplexed data transmissions is typically performed on the reference symbols 308 in the frequency domain, because it is computationally more efficient to do so. The resulting channel estimates 310 are then used to extract 312 the transmitted symbols 314 from the message data 316.
A joint channel estimation in the time domain instead over the full bandwidth of the received signal 300 would perform better than the frequency domain channel estimation (“FDCE”) calculation 306. Current approaches for performing time domain joint channel estimations, however, have been computationally infeasible in practice.
What is needed, therefore, is an improved, computationally feasible method and system for performing time domain joint channel estimations on a frequency-multiplexed data transmission comprising a plurality of simultaneous signals originating from a plurality of signal sources.